Вы находитесь на странице: 1из 71

University of Tennessee, Knoxville

Trace: Tennessee Research and Creative


Exchange
Masters Theses Graduate School

5-2014

Microgrid Modeling and Grid Interconnection


Studies
Hira Amna Saleem
University of Tennessee - Knoxville, hsaleem1@utk.edu

Recommended Citation
Saleem, Hira Amna, "Microgrid Modeling and Grid Interconnection Studies. " Master's Thesis, University of Tennessee, 2014.
http://trace.tennessee.edu/utk_gradthes/2756

This Thesis is brought to you for free and open access by the Graduate School at Trace: Tennessee Research and Creative Exchange. It has been
accepted for inclusion in Masters Theses by an authorized administrator of Trace: Tennessee Research and Creative Exchange. For more information,
please contact trace@utk.edu.
To the Graduate Council:
I am submitting herewith a thesis written by Hira Amna Saleem entitled "Microgrid Modeling and Grid
Interconnection Studies." I have examined the final electronic copy of this thesis for form and content
and recommend that it be accepted in partial fulfillment of the requirements for the degree of Master of
Science, with a major in Electrical Engineering.
Kai Sun, Major Professor
We have read this thesis and recommend its acceptance:
Fred Wang, Fran Li
Accepted for the Council:
Dixie L. Thompson
Vice Provost and Dean of the Graduate School
(Original signatures are on file with official student records.)
Microgrid Modeling and Grid Interconnection Studies

A Thesis Presented for the


Master of Science
Degree
The University of Tennessee, Knoxville

Hira Amna Saleem


May 2014
Copyright 2014 by Hira Amna Saleem
All rights reserved.

ii
ACKNOWLEDGEMENTS
I would like to first thank my advisor Dr. Kai Sun for giving me this opportunity and for guiding
me throughout the project. This work is partially from an ongoing NEC project titled Novel
Control Techniques for Enhancement of Microgrid Stability in the Grid Mode. I would also like
to express my appreciation to NEC for supporting this project. The other contributors of the
project are Yongli Zhu and Riyasat Azim. I would also like to thank Dr. Fran Li and Dr. Fred
Wang for serving on my thesis committee.

iii
DEDICATION

To my husband,
Ahmad

To my mother,
Sabiha

iv
ABSTRACT

The demand for renewable energies and their integration to the grid has become more pressing
than ever before due to the various reasons including increasing population energy demand,
depleting fossil fuels, increasing atmospheric population, etc. Thus the vision of a sustainable
future requires easy and reliable integration of renewable distributed generators to the grid. This
masters thesis studies the dynamics of distributed generators when they are connected with the
main grid. Simulink MATLAB is used for the design and simulations of this system. Three
distributed generators are used in this system: Photo-voltaic converter, Fuel cell and diesel
generator. The control and design of the power electronics converters is done to function
properly in both grid-connected and islanding mode. The turbine governors in diesel generators
control the proper functioning of diesel generator in both modes. The converters in both battery
and PV make sure that they work properly in both grid-connected and islanding mode. The
control of battery converter is designed in a way to function for load-shaving during unplanned
load changes in the microgrid. This fully functioning microgrid is then connected with the main
grid using Kundurs two-area system and simulated for various faults and load changes. A
collection of data at the point of common coupling which is the point of connection of microgrid
and main grid is gathered for various cases in the grid-connected mode. The cases for faults in
the external grid are simulated and then WEKA software is used to develop decision trees. The
development of the decision trees can help in predicting the decision of islanding of microgrid.
By increasing this database for more scenarios; the response of the generators in grid and
distributed generators in microgrid can be studied with decision trees giving more accurate
results.

v
TABLE OF CONTENTS

Chapter 1 : INTRODUCTION........................................................................................................ 1
Distributed Generation ................................................................................................................ 1
Microgrid .................................................................................................................................... 1
Types of fault .............................................................................................................................. 2
Modeling of Microgrid ............................................................................................................... 2
IEEE 13 Node Test Feeder System ............................................................................................. 2
Chapter 2 : DIESEL GENERATOR ............................................................................................... 6
Design and Modeling .................................................................................................................. 6
Hydraulic Turbine Governor (Grid-mode) ............................................................................. 6
Diesel Engine Governor (Islanding-mode) ............................................................................. 7
Transformer: ........................................................................................................................... 7
Simulink Model ...................................................................................................................... 7
Chapter 3 : PHOTOVOLTAIC ENERGY SOURCE ..................................................................... 8
Importance of PV in energy production...................................................................................... 8
Design and Modeling .................................................................................................................. 8
Maximum Power Point Tracking (MPPT): ............................................................................ 8
Average model of DC-DC converter and three-phase inverter: ............................................. 9
Active and Reactive Power control of converter: ................................................................... 9
Transformer: ........................................................................................................................... 9
Simulink Model ...................................................................................................................... 9
Chapter 4 : BATTERY ................................................................................................................. 10
Lead-Acid Battery................................................................................................................. 10
Design and Modeling ................................................................................................................ 10
Bi-directional Converter: ...................................................................................................... 11
Real/reactive Power Controller: ............................................................................................ 12
LCL Filter: ............................................................................................................................ 12
Transformer: ......................................................................................................................... 13
Simulink Model .................................................................................................................... 13
Chapter 5 : CONTROL STRATEGIES IN MICROGRID ........................................................... 14
Islanding Mode ......................................................................................................................... 14
Grid-Connected Mode .............................................................................................................. 15
IEEE 1547 Standard Adherence ............................................................................................... 17
Current Harmonics: ............................................................................................................... 17
DC Injection: ......................................................................................................................... 19
Chapter 6 : SIMULATIONS AND RESULTS ............................................................................. 20
Grid Interconnection ................................................................................................................. 20
1. Single-phase 4-cycles fault at PCC ............................................................................... 21
2. Three-phase 4-cycles fault at Bus 632 near PV and PV is turned off during fault. ...... 24
3. Sudden load decrease in microgrid ............................................................................... 27
Kundurs Two Area System...................................................................................................... 30
Three phase 10-cycles fault in grid-mode at Bus 1 followed by islanding:.......................... 32
Chapter 7 : DECISION TREES .................................................................................................... 36
Chapter 8 CONCLUSION AND RECOMMENDATIONS ......................................................... 54

vi
LIST OF REFERENCES .............................................................................................................. 56
VITA ............................................................................................................................................. 59

vii
LIST OF TABLES

Table 1.1 Overhead Line Configuration Data................................................................................. 3


Table 1.2 Underground Line Configuration Data. .......................................................................... 3
Table 1.3 Line Segment Data.......................................................................................................... 4
Table 1.4. Transformer Data. .......................................................................................................... 4
Table 1.5. Capacitor Data ............................................................................................................... 4
Table 1.6. Regulator Data. .............................................................................................................. 4
Table 1.7. Spot Load Data. ............................................................................................................. 5
Table 1.8. Distributed Load Data. ................................................................................................... 5
Table 5.1. Current Harmonics Standard in IEEE 1547. ................................................................ 17
Table 5.2. DC Current Injected by DGs. ..................................................................................... 19
Table 7.1. Interconnection System Response to Abnormal Voltages [15] ................................... 41
Table 7.2. Interconnection System Response to Abnormal Frequencies [15] ............................. 41
Table 7.3. Interconnection System Response to Abnormal Voltages with Class Identifier ......... 42
Table 7.4. Interconnection System Response to Abnormal Frequencies with Class Identifier .... 42
Table 7.5. List of critical attributes selected for building the decision trees ............................... 44
Table 7.6. Decision tree Training Data for Frequency Classification Problem ............................ 45
Table 7.7. Decision tree Training Data for Voltage Classification Problem ................................ 46
Table 7.8. Ranking of Predictor Variables and Corresponding Accuracy in Frequency
Classification Problem .......................................................................................................... 47
Table 7.9. Ranking of Predictor Variables and Corresponding Accuracy in Voltage Classification
Problem ................................................................................................................................. 48

viii
LIST OF FIGURES
Figure 1.1. Network topology of IEEE 13 Node Test Feeder System. ........................................... 3
Figure 1.2. Simulink Model of IEEE 13 Node Test Feeder System. .............................................. 5
Figure 2.1. Hydraulic Turbine Governor model in Simulink MATLA .......................................... 6
Figure 2.2. Simulink block of Diesel Engine Governor ................................................................. 7
Figure 2.3. Simulink model of Diesel Generator. ........................................................................... 7
Figure 3.1. Simulink model of PV. ................................................................................................. 9
Figure 4.1. Simulink model of lead-acid battery with its parameters ........................................... 11
Figure 4.2. Bi-directional Voltage Source Converter ................................................................... 11
Figure 4.3. P-Q Controller for battery .......................................................................................... 12
Figure 4.4. LCL filter for batter .................................................................................................... 13
Figure 4.5. Simulink model of Battery ......................................................................................... 13
Figure 5.1. Simulink block of Diesel Engine Governor ............................................................... 14
Figure 5.2. PQ- control design [16]. ............................................................................................. 14
Figure 5.3. Hydraulic Turbine Governor model in Simulink MATLAB ..................................... 15
Figure 5.4. Microgrid Power flow diagram .................................................................................. 16
Figure 5.5. Battery charge and discharge control for power shaving. .......................................... 16
Figure 5.6. Power levelling by using battery to discharge for 100kW load increase of load in
microgrid. .............................................................................................................................. 17
Figure 5.7. Current Harmonics injected by Diesel Generator. ..................................................... 18
Figure 5.8. Current Harmonics injected by PV............................................................................. 18
Figure 5.9. Current Harmonics injected by battery....................................................................... 19
Figure 6.1. Simulink model of Microgrid. .................................................................................... 20
Figure 6.2. Single-phase 4-cycles fault at PCC in Simulink model of Microgrid. ....................... 21
Figure 6.3. V, I, P and Q at PCC................................................................................................... 22
Figure 6.4.V, I, P and Q of Battery. ............................................................................................ 22
Figure 6.5. V, I, P and Q of PV.................................................................................................... 23
Figure 6.6. V, I, P and Q of Diesel Generator. ........................................................................... 23
Figure 6.7. Frequencies at PCC, battery, PV and Diesel generator respectively. ......................... 24
Figure 6.8. Three-phase 4-cycles fault at Bus 632 in Simulink model of Microgrid. .................. 24
Figure 6.9. V, I, P and Q at PCC................................................................................................... 25
Figure 6.10. V, I, P and Q of Battery ............................................................................................ 25
Figure 6.11. V, I, P and Q of PV................................................................................................... 26
Figure 6.12. V, I, P and Q of Diesel Generator. ........................................................................... 26
Figure 6.13. Frequencies at PCC, battery, PV and Diesel generator respectively. ....................... 27
Figure 6.14. Sudden load decrease in microgrid........................................................................... 27
Figure 6.15. V, I, P and Q at PCC................................................................................................ 28
Figure 6.16. V, I, P and Q of Battery. .......................................................................................... 28
Figure 6.17. V, I, P and Q of PV................................................................................................. 29
Figure 6.18. V, I, P and Q of Diesel Generator. .......................................................................... 29
Figure 6.19. Frequencies at PCC, battery, PV and Diesel generator respectively. ...................... 30
Figure 6.20. Tie-line diagram of Kundurs Two-Area System [21]. ............................................ 30
Figure 6.21(a) Simulink Model of Kundurs Two-Area System .................................................. 31
Figure 6.22. Active Power Flow at PCC, Battery, PV and Diesel Generator. .............................. 32

ix
Figure 6.23. Reactive Power Flow at PCC, Battery, PV and Diesel Generator. ........................ 33
Figure 6.24. Frequencies of generators in Kundurs Two-Area System. ..................................... 33
Figure 6.25. Frequencies at PCC, battery, PV and Diesel generator. .......................................... 34
Figure 6.26. V, I, P and Q of Battery ............................................................................................ 34
Figure 6.27. V, I, P and Q of PV................................................................................................... 35
Figure 6.28. V, I, P and Q of Diesel Generator in microgrid...................................................... 35
Figure 7.1. A simple 5 node decision tree. .................................................................................. 37
Figure 7.2. DT based scheme for microgrid islanding and controls ............................................. 39
Figure 7.3. Decision trees for voltage classification: (a) with predictors Ibmin, Icmin, Ibmax; ......
(b) with predictors Ibmax and Icmin. ................................................................................... 50
Figure 7.4. Decision trees for voltage classification: (a) with predictors Icmin, Pmax and
Qmax; (b) with predictors Qmax and Iamax. ....................................................................... 50
Figure 7.5. Decision trees for frequency classification: (a) with predictors Icmin and Vamin;..
(b) with predictors Iamax and Icmin. .................................................................................... 51
Figure 7.6. Decision trees for frequency classification: (a) with predictors Qmax and Vamin;..
(b) with predictors Vcmax and Vbmin. ................................................................................ 51
Figure 7.7. Decision trees for frequency classification with predictors Icmax, Vamin and.
Fault types. ............................................................................................................................ 52
Figure 7.8. Decision trees for frequency classification with predictors Vamax, Pmin & Pmax. . 52
Figure 7.9. Decision trees for voltage classification with predictors Iamin, Pmin & Pmax. ........ 53
Figure 7.10. Decision trees for voltage classification with predictors Fault types, Iamax, Pmin
and Ibmax.............................................................................................................................. 53

x
CHAPTER 1 : INTRODUCTION

Distributed Generation

The worldwide increase in demand of energy necessitates the use of different electric energy
sources combined by a common grid for making an efficient electrical energy network. The
distributed generation is becoming more and more popular compared to the conventional
centralized system. In [1], the world scenarios are discussed as how the increasing demand of
energy can be compensated. WADE is a non-profit organization which surveyed and then
analyzed the energy profile of different countries and consequently, stressed the need of the full
utilization of Distributed Electricity generation in order to keep up with the worldwide growing
demand of energy. The share of decentralized power generation in the world market has
increased to 7.2% by 2004, up from 7% in 2002. The long discussed and expected transition
from a central power model to a hybrid DE-central mix may possibly be underway, though
slowly. WADE is optimistic that this market share will continue to expand [1].

The generation of highly reliable, good quality electrical power near the place where it is
demanded can imply a change of paradigm. This concept, named distributed generation (DG), is
especially promising when dispersed energy storage systems (fuel cells, compressed-air devices,
or flywheels) and renewable energy resources (photovoltaic arrays, variable speed wind turbines,
or combined cycle plants) are available. These resources can be connected through power
conditioning ac units to local to local electric power networks also known as microgrids [2]. The
distributed sources connection done with the grid should be in compliance with the IEEE 1547
standards. The penetration of distributed energy sources in the main grid is increasing and with it
the need for the development of better and more reliable protection schemes is also increasing.

The few main advantages of distributed generation are given below:

1. The transmission and distribution cost is lowered.


2. Distributed Generation promotes the use of alternative energy sources and reliance on
fossil fuels and nuclear power is reduced.
3. There is higher fuel efficiency because localized generation allows use of heat as well as
generating electricity.
4. Distributed Energy Sources require less backup (as outage of a single generator cannot
impact a system of many small generators), unlike a central system consisting of many
large power plants. So, there are lesser chances of total black-out.

Microgrid

The DOE definition of microgrid is given below:

A group of interconnected loads and distributed energy resources (DER) with clearly
defined electrical boundaries that acts as a single controllable entity with respect to the
grid (and can) connect and disconnect from the grid to enable it to operate in both grid-
connected or island mode.

1
The heart of the microgrid concept is the notion of a flexible, yet controllable electronic interface
between the microgrid and the familiar wider power system, or macro-grid. This interface
essentially isolates the two sides electrically; and yet connects them economically by allowing
delivery and receipt of electrical energy and ancillary services at the interface [5].The
consideration of bidirectional power flow between microgrid and main grid puts forward many
challenges in terms of protection and reliability.

Types of fault
Most faults in an electrical utility system with a network of overhead lines are one-phase-to-
ground faults resulting primarily from lightning-induced transient high voltage and from falling
trees and tree limbs. Momentary tree contact caused by wind, ice, freezing snow, and wind
during severe storms are other causes of fault. These faults include the following, with very
approximate percentages of occurrence [3]:

Single phase-to-ground: 70%80%


Phase-to-phase-to ground: 17%10%
Phase-to-phase: 10%8%
Three-phase: 3%2%

In conventional network, power flows from higher voltage level to lower voltage level and in
case of fault short circuit current decrease as distance increase. Modern Microgrid has been
changed the concept and power could flow in both direction. Some of the prominent protection
issues are: short circuit power, fault current level and direction, device discrimination, reduction
in reach of over-current relays, nuisance tripping, protection blinding, etc [4].

Modeling of Microgrid

Simulink MATLAB is used to model microgrid in this project. Three distributed generators
(Lead-acid battery, diesel generator and PV) are used in this microgrid. The location of the
distributed generators is chosen to be close to each other so that they can support each other in
functioning.

IEEE 13 Node Test Feeder System

A benchmarked radial distribution was used to develop the load and lines model of microgrid
model to help in comparative studies. IEEE 13 Node test feeder system shown in Figure 1.1 was
used for this purpose. Some of the characteristics of this system are given below:

1. It is small but heavily loaded at 4.16kV.


2. It has unbalanced spot and distributed loads.
3. It consists of lines with unbalanced phasing.
4. Shunt capacitor banks.

2
Bus 650 is the interconnection point of microgrid and main grid and is called Point of Common
Coupling (PCC). Power can flow in both directions at this point.

Figure 1.1. Network topology of IEEE 13 Node Test Feeder System.

The details of this IEEE 13 Node test feeder model as specified by IEEE in [6] are given below:
Table 1.1 Overhead Line Configuration Data.
Config. Phasing Phase Neutral Spacing
ACSR ACSR ID
601 BACN 556,500 26/7 4/0 6/1 500
602 CABN 4/0 6/1 4/0 6/1 500
603 CBN 1/0 1/0 505
604 ACN 1/0 1/0 505
605 CN 1/0 1/0 510

Table 1.2 Underground Line Configuration Data.


Config. Phasing Cable Neutral Space ID
606 ABCN 250,000 AA, CN None 515
607 AN 1/0 AA, TS 1/0 Cu 520

3
Table 1.3 Line Segment Data.
Node A Node B Length(ft.) Config.
632 645 500 603
632 633 500 602
633 634 0 XFM-1
645 646 300 603
650 632 2000 601
684 652 800 607
632 671 2000 601
671 684 300 604
671 680 1000 601
671 692 0 Switch
684 611 300 605
692 675 500 606

Table 1.4. Transformer Data.


kVA kV-high kV-low R-% X-%
Substation: 5,000 115 - D 4.16 Gr. Y 1 8
XFM -1 500 4.16 Gr.W 0.48 Gr.W 1.1 2

Table 1.5. Capacitor Data


Node Ph-A Ph-B Ph-C
kVAr kVAr kVAr
675 200 200 200
611 100
Total 200 200 300

Table 1.6. Regulator Data.


Regulator ID: 1
Line Segment: 650 - 632
Location: 50
Phases: A - B -C
Connection: 3-Ph,LG
Monitoring Phase: A-B-C
Bandwidth: 2.0 volts
PT Ratio: 20
Primary CT Rating: 700
Compensator Settings: Ph-A Ph-B Ph-C
R - Setting: 3 3 3
X - Setting: 9 9 9
Volltage Level: 122 122 122

4
Table 1.7. Spot Load Data.
Node Load Ph-1 Ph-1 Ph-2 Ph-2 Ph-3 Ph-3
Model kW kVAr kW kVAr kW kVAr
634 Y-PQ 160 110 120 90 120 90
645 Y-PQ 0 0 170 125 0 0
646 D-Z 0 0 230 132 0 0
652 Y-Z 128 86 0 0 0 0
671 D-PQ 385 220 385 220 385 220
675 Y-PQ 485 190 68 60 290 212
692 D-I 0 0 0 0 170 151
611 Y-I 0 0 0 0 170 80
TOTAL 1158 606 973 627 1135 753

Table 1.8. Distributed Load Data.


Node A Node B Load Ph-1 Ph-1 Ph-2 Ph-2 Ph-3 Ph-3
Model kW kVAr kW kVAr kW kVAr
632 671 Y-PQ 17 10 66 38 117 68

The total load at all three phases in the microgrid is unbalanced and is given below:
Active Power:
A-phase: 1.175 MW
B-phase: 1.039 MW
C-phase: 1.625 MW
Reactive Power:
A-phase: 416 kVAR
B-phase: 465kVAR
C-phase: 878kVAR
The IEEE 13 Node Test Feeder Model developed in MATLAB Simulink is shown in Figure 1.2.

Figure 1.2. Simulink Model of IEEE 13 Node Test Feeder System.

5
CHAPTER 2 : DIESEL GENERATOR
A diesel generator consists of a diesel engine with an electric generator to generate electrical
energy. A diesel generator is chosen to be installed in the microgrid installed; it will be of much
higher power capacity than battery and PV and will support the grid.

Emergency standby diesel generators, for example such as those used in hospitals, water plant,
are, as a secondary function, widely used in the US and the UK (Short Term Operating Reserve)
to support the respective national grids at times for a variety of reasons. In the UK for example,
some 0.5 GW of diesels are routinely used to support the National Grid, whose peak load is
about 60 GW. These are sets in the size range 200 kW to 2 MW. This usually occurs during, for
example, the sudden loss of a large conventional 660 MW plant, or a sudden unexpected rise in
power demand eroding the normal spinning reserve available [7].

Design and Modeling

A three-phase generator rated 3.125 MVA, 2.4kV is connected to a 4.16 kV network through a
Delta-Y 5 MVA transformer.

Hydraulic Turbine Governor (Grid-mode)


In grid connected mode, operational control of voltage and frequency is done entirely by the
grid; however, microgrid still supplies the critical loads at PCC, thus acts as a PQ bus. The
electricity from the main grid is used as a reference and the hydraulic turbine governor of
generator maintains and follows this frequency. The Hydraulic Turbine and Governor block in
Simulink uses a nonlinear hydraulic turbine model, a PID governor system, and a servomotor as
shown in Fig.2.1.

Figure 2.1. Hydraulic Turbine Governor model in Simulink MATLA

6
Diesel Engine Governor (Islanding-mode)
In islanding mode, there is no reference signal from the main grid. In this microgrid model,
generator provides the reference signal in the microgrid and controls the voltage and frequency
using the diesel engine governor. The Simulink model of the Diesel engine governor implements
the actuator control as shown in Fig.2.2.

Figure 2.2. Simulink block of Diesel Engine Governor

Transformer:
A step-up transformer of 5 MVA is used to step up the voltage from the diesel generator to the
voltage level of 4.16kV in the microgrid. The transformer steps up voltage from 2.4 kV to
4.16kV.

Simulink Model
The Simulink model used is show below in Fig.2.3.

Figure 2.3. Simulink model of Diesel Generator.

7
CHAPTER 3 : PHOTOVOLTAIC ENERGY SOURCE
The source of photovoltaic energy is sun which is 150 million kilometer away from Earth. It is
such a big source of energy that in one minute, the Earth receives enough energy from the Sun to
meet our needs for a year. It is also easily accessible and free of cost. The solar energy can be
used directly for heating, lighting, cooking, drying and for generating electricity. On a large
scale, this solar energy can be utilized for commercial and industrials use aswell. Winds and
waves are also consequences of the solar energy. The thermal energy from sun is converted to
electricity using solar cells. The direct current (DC) electricity created in this way is then
converted to alternating current (AC) or stored for later use.

Importance of PV in energy production

Photovoltaics is a fast growing market and holds a promising future. Worldwide in 2011 about
half of the previously cumulated PV module capacity entered the market. Now PV technology is
being increasingly recognized as a part of the solution to the growing energy challenge and an
essential component of future global energy production. PV system performance has strongly
improved. The typical Performance Ratio has increased from 70 % to about 85 % over the last 15
years. Germany including other European country contributed to major role towards the global
cumulative PV installation until 2011, that is, 70% of global PV installation [8].

Design and Modeling

PV model of 0.6 MW capacity is designed in Simulink MATLAB. Average models of converter


are used to emulate the converter behavior which is controlled by PQ-Control strategy and then
the output is stepped up to the network voltage level of 4.16kV.

Maximum Power Point Tracking (MPPT):


Irradiance data from solar radiation is input to output the DC voltage based on the Maximum
Power Point Tracking (MPPT).

Power output of a Solar PV module changes with change in the direction of sun, changes in solar
irradiance level and variations in temperatures. It is known that the efficiency of the solar PV
module is low and it is in the range of 13%. Since, the module efficiency is low it is desirable to
operate the module at the peak power point so that the maximum power can be delivered to the
load under varying temperature and irradiance conditions. Hence maximization of power,
improves the utilization of the solar PV module. A maximum power point tracker (MPPT) is
used for extracting the maximum power from the solar PV module and transferring that power to
the load. Considering the investment cost of the PV system, it is always a prerequisite to operate
PV at its Maximum Power Point (MPP) [9].

8
Average model of DC-DC converter and three-phase inverter:
After the conversion of solar energy to DC, an average model of DC-DC converter is used to
step up the voltage. This stepped up DC voltage is then converted to alternating current (AC) by
an average model of three-phase inverter.

An inductor of 6.74H is used to reduce the current harmonics at the output of the PV converter.

Active and Reactive Power control of converter:


Current controlled based PQ Control Strategy is used to provide reference d and q
components which are then input to the average model of converter. The reference reactive
power in PV is a constant value in grid-connected mode but in islanding mode, a droop control is
used to follow the reactive power in the islanded microgrid.

Transformer:
A step-up transformer of 0.5 MVA is used to step up the voltage of PV to the voltage level of
4.16kV in the microgrid. The transformer steps up voltage from 0.48kV to 4.16kV.

Simulink Model
The Simulink model used is show below in figure.

Figure 3.1. Simulink model of PV.

9
CHAPTER 4 : BATTERY
Fuel Cells are a popular distributed energy source. They can be used for hospitals, offices,
residential areas or to supply power at remote places like construction areas. Lead-acid batteries
have comparatively lower energy density but they are cheaper and hence, more suitable for large
systems where weight is not a concern.

Lead-Acid Battery
Renewable energy systems like Solar and Wind are intermittent in nature. For stand-alone
applications they often require energy storage systems to provide the fill in power. Batteries
have been traditionally used to provide the fill-in power for solar and wind systems
The Lead Acid Battery is one of the widely used electrochemical energy storage systems. This
can be attributed to its chemical and physical properties that makes it an ancient system and
suitable for a variety of applications [22]. A few of these properties are given below [22].

The reactants are solids of low solubility which causes a stable voltage and highly
reversible reactions.
Both electrodes contain only lead and lead compounds as active material that does not
require conducting additives.
It has a high cell voltage of 2V.
Lead Acid technology is cheaper than most technologies and is the primary reason.

A lead acid battery is used to compensate the ups and downs in the grid-tied voltage during grid-
connection mode. In islanding mode, the same battery is used to fulfill the load requirements. In
[2], a 34 MW NAS Battery is used to stabilize the output of a 51 MW Wind Turbine farm in the
Tahoku district of Japan. The fluctuations in the power supplied by the wind turbine are
stabilized with a precision of +-2% using the battery.

Design and Modeling

The simulation of charge and discharge of a battery model using parameters taken from a steady
state curve is done in [10]. The approximate error between simulation results and experimental
results varies between 5% to 10% [10]. The battery model block in Simulink is used to get the
simulation results in [10]. The battery block in Simulink implements set of predetermined charge
behavior for four types of battery: Lead-Acid, Lithium-Ion, Nickel-Cadmium and Nickel-Metal-
Hydride. The Lead-Acid Battery Model is therefore used to emulate the exact behavior of a real
lead-acid battery. The lead-acid battery used with its parameters is shown in Fig.4.1

10
Figure 4.1. Simulink model of lead-acid battery with its parameters

Bi-directional Converter:
Bidirectional converter is used in electric vehicles for efficient utilization of battery by providing
a uniform voltage supply and re-charging the battery at some instances e.g down-slope motion of
car [12-14]. A bidirectional voltage source converter shown in Fig.4.2 is used in the battery
converter design. This converter acts as a buck converter when the battery discharges and the
current flows from battery to the main grid. Otherwise, it acts a boost converter when the battery
charges and the current flows from the main grid to the battery.

Figure 4.2. Bi-directional Voltage Source Converter

11
Real/reactive Power Controller:
The main purpose of this controller is to control the instantaneous real and reactive power that
this battery exchanges with the microgrid side. The flow and direction of the current is controlled
by controlling the PWM given to this converter. The total power for which the converter is
designed is 0.15 MW and the lead-acid battery of 1088 V is selected. Switching frequency of
8kHz is used to generate the PWM for the converter; the PWM is generated using the dq- frame
of reference. PQ control strategy is used for this purpose.

Figure 4.3. P-Q Controller for battery

In this microgrid connected VSC system, the real and reactive power output of the battery
converter is proportional to the d- and q-axis components of the converter current, respectively.
The angle of the grid-connected side is calculated using the phase-locked loop (PLL) of voltage
at that point and provided to the controller.

LCL Filter:
Harmonic current is injected by the voltage source converter used which is controlled by PWM
switching. Therefore, an AC filter is used on the grid side of the converter to filter the high
frequency harmonics injected by the converter. An LCL filter as shown in Figure 4.4 is used to
remove these harmonics.

12
Figure 4.4. LCL filter for batter

Transformer:
A step-up transformer is used to step up the voltage of the battery to the voltage level of 4.16kV
in the microgrid. The transformer steps up voltage from 220V to 4.16kV.

Simulink Model
The Simulink model used is show below in Figure 4.5.

Figure 4.5. Simulink model of Battery

13
CHAPTER 5 : CONTROL STRATEGIES IN MICROGRID
The control of a microgrid to ensure a reliable and stable supply is very important for the safety
and stability of main grid. The microgrid model made can work in both grid-connected and
islanding mode. The operation in both the modes is controlled with the help of convertor
controllers in each distributed generator. The control strategies used in both modes are explained
below.

Islanding Mode

In islanding mode, there is no reference signal from the main grid. The microgrid operates
independently and acts like a PV bus with hydraulic turbine governor controlling the frequency.
In this microgrid model, generator provides the reference signal in the microgrid and controls the
voltage and frequency using the diesel engine governor. The Simulink model of the Diesel
engine governor implements the actuator control as shown in Fig.5.1.

Figure 5.1. Simulink block of Diesel Engine Governor


Battery and PV operate in PQ mode; the PQ control design explained in [16] is used and is
shown below.

Figure 5.2. PQ- control design [16].

This PQ control consists of an inner current control loop and the control structure contains an
inner and an outer voltage control loop. The control mechanism is performed in dq reference
frame rotating at the fundamental frequency. This fundamental frequency is calculated form the
PLL using the voltage and current at the output of the converter for Parks Transformation. The
benefit of using the dq reference frame is that the sinusoidal command tracking problem is

14
converted to an equivalent DC command tracking problem. Therefore, PI compensators can be
used in the control design. Also, by using integral in compensator an almost zero steady-state
tracking error can be attained.

The reference reactive power in PV is a constant value in grid-connected mode but in islanding
mode, a droop control is used to follow the reactive power in the islanded microgrid.

Grid-Connected Mode

In grid connected mode, operational control of voltage and frequency is done entirely by the
grid; however, microgrid still supplies the critical loads at PCC, thus acts as a PQ bus. The
electricity from the main grid is used as a reference and the hydraulic turbine governor of
generator maintains and follows this frequency. The Hydraulic Turbine and Governor block in
Simulink uses a nonlinear hydraulic turbine model, a PID governor system, and a servomotor as
shown in Fig.5.3.

Figure 5.3. Hydraulic Turbine Governor model in Simulink MATLAB

Battery and PV operate in PQ mode ; the abc- reference frame in these controllers is converter to
dq reference frame. The signals in steady-state are turned to DC waveforms which can be
controlled more easily using PI controllers. The PQ designed controller used by battery and PV
is based on the Fig.5.2.

Battery control for Power-Shaving


The battery is used as a compensator to minimize the variation of the PCC MW flow relative to
the planned amount. Fig.5.4 shows the flow of power throughout the microgrid.

15
Figure 5.4. Microgrid Power flow diagram
Battery control is designed to shave power above and below a reference value for both grid-
connected and islanding mode.

That is minimizing,
= - (5.1)

If , battery is charged. Otherwise, battery is discharged. For this purpose, an additional


control is added to the PQ controller design of battery as shown in Fig.5.5.

Figure 5.5. Battery charge and discharge control for power shaving.

The error which is the difference of the active and reactive power flow through the tie-line and
the planned flow is passed through a compensator and then added to the desired battery power

16
which we want it to operate at and thus give the reference active and reactive power for the PQ-
controller of battery. Also, the active and reactive power of battery is subtracted from that
through tie-line so that the power produced by battery does not affect the output reference power
of controller.

For 100kW load increase in microgrid,

Figure 5.6. Power levelling by using battery to discharge for 100kW load increase of load in microgrid.

From the Fig.5.6, there is substantial levelling of the active power at PCC by using battery when
there is an abnormal load shift in microgrid. In this way, the battery can be used to shave peak
load change. Whenever the power flow through PCC changes from the planned value, battery
will charge and discharge accordingly to minimize the impact of the unplanned load on the
power flow at PCC.

IEEE 1547 Standard Adherence

The 1547 standard is the only systems-level technical standard of uniform requirements and
specifications universally needed to interconnect distributed energy resources with the grid.

Current Harmonics:
According to IEEE 1547 Standard, when the DR is serving balanced linear loads, harmonic
current injection at PCC shall not exceed the limits stated below [15].
Table 5.1. Current Harmonics Standard in IEEE 1547.

17
The harmonics injected by the diesel generator satisfy the limitations stated in table 5.1 and are
shown in the Fig.5.7 below:

Figure 5.7. Current Harmonics injected by Diesel Generator.

The harmonics injected by PV converter satisfy the limitations stated in table 5.1 and are shown
in the Fig.5.8 below:

Figure 5.8. Current Harmonics injected by PV.

The harmonics injected by the battery converter satisfy the limitations stated in Table 5.1 and are
shown in the figure below:

18
Figure 5.9. Current Harmonics injected by battery.

DC Injection:
According to the limitation of dc injection in IEEE Standard 1547-2003[15]:

The DR and its interconnection system shall not inject dc current greater than 0.5% of
the full rated output current at the point of DR connection.

These limitations are satisfied by all three Distributed Generators as shown in the Table 5.2
below.

Table 5.2. DC Current Injected by DGs.

Distributed Generators DC Injection Current (A)


Battery 0.0178
PV 0.137
Generator 5.8
Total DC Injection 5.95 <6.5 (0.5% of full rated output
Current current)

19
CHAPTER 6 : SIMULATIONS AND RESULTS

Grid Interconnection
The motivation and benefits of connecting the distributed generators to main grid are given
below:

1. Stable Operation:
The power grid acts a stiff source and facilitates stable operation of microgrid.
2. Availability:
Grid is available easily to take care of load in microgrid in cases when the distributed
generators in microgrid may go down.
3. Economics:
The extra electricity generated by microgrid can be sent to grid and sold for economic
benefits. Also, electricity from grid can be used at times like night when the output of PV is low
and the price of electricity is cheaper at grid.

An ideal voltage source can be used as grid in Simulink. The three DGs are placed closer to each
other so that they operate properly as shown in Figure 6.1.

Figure 6.1. Simulink model of Microgrid.

20
The following three contingencies simulations are done in grid-mode in order to study and
analyze the behavior of the microgrid.

1. IEEE 13 Feeder Microgrid with single-phase 4-cycles fault at PCC.


2. IEEE 13 Feeder Microgrid with three-phase 4-cycles fault at Bus 633 near PV and PV is
turned off during fault.
3. IEEE 13 Feeder Microgrid with sudden load change.

1. Single-phase 4-cycles fault at PCC


The 4-cycles fault is introduced at PCC as shown in Figure 6.2.

Figure 6.2. Single-phase 4-cycles fault at PCC in Simulink model of Microgrid.

The voltage, current, active power and reactive power simulation results at PCC are shown in
Fig.6.3.The PCC voltage attains its steady state value in about 0.1 second after fault. There is a
huge dip in the reactive power but the active power remains almost constant. Since its a single
phase fault, the variations in the voltage and current of all three phases are different during the
duration of fault. The active and reactive power at PCC is flowing from the main grid to
microgrid but during the duration of fault the flow of the reactive power reverses.

21
Figure 6.3. V, I, P and Q at PCC.
The voltage, current, active power and reactive power simulation results of battery are shown in
the figure below. Battery current jumps from 12A to 4800A as shown in the figure. The voltage
of two phases changes from 6000V to 2500V approximately. The reactive power supplied by the
battery increases during the duration of fault and is contributes to being supplied to the main grid
at PCC.

Figure 6.4.V, I, P and Q of Battery.

22
The voltage, current, active power and reactive power simulation results of PV are shown below.
The current increases during the fault but gradually attains its normal case value.

Figure 6.5. V, I, P and Q of PV.

The voltage, current, active power and reactive power simulation results of Diesel Generator are
shown below. There is a huge increase in the current of one phase during the duration of fault.
Also, the reactive power increases during the duration of fault and contributes to the reactive
power being supplied during that interval to main grid.

Figure 6.6. V, I, P and Q of Diesel Generator.

23
The frequencies at PCC, battery, PV and Diesel in the given order are shown below. The
frequencies attain their stable value of 60Hz after the fault clear in about 0.5 seconds.

Figure 6.7. Frequencies at PCC, battery, PV and Diesel generator respectively.

2. Three-phase 4-cycles fault at Bus 632 near PV and PV is turned off during fault.
The 4-cycles three-phase fault is introduced at Bus 632 as shown in Figure 6.8.

Figure 6.8. Three-phase 4-cycles fault at Bus 632 in Simulink model of Microgrid.

24
The voltage, current, active power and reactive power simulation results at PCC are shown in
Figure 6.9. The flow of power is from microgrid to main grid at PCC. However, during the fault
the active power flow reverses at PCC as shown below.

Figure 6.9. V, I, P and Q at PCC


The voltage, current, active power and reactive power simulation results of battery are shown in
Figure 6.10.Battery being away from the point of three-phase fault is much less affected. The
battery functions in its normal state except a sharp current peak at the time of fault.

Figure 6.10. V, I, P and Q of Battery

25
The voltage, current, active power and reactive power simulation results of PV are shown below
in Figure 6.11.During the fault, PV disconnects from the microgrid so the current and power
drops to zero.

Figure 6.11. V, I, P and Q of PV.

The voltage, current, active power and reactive power simulation results of Diesel Generator are
shown below. During the fault, the voltage drops to a lower value and current increases. There is
some variation in the active and reactive power flow from the diesel generator and it does not
attain its stable value even after the fault clears and become stable after about 0.3 seconds.

Figure 6.12. V, I, P and Q of Diesel Generator.

26
The frequencies at PCC, battery, PV and Diesel generator in the given order are shown
below.The frequency of PV gets higher after disconnection but that is irrelevant as it wont be
connected to the microgrid.The system is well synchronized at 60 Hz after the fault clears.

Figure 6.13. Frequencies at PCC, battery, PV and Diesel generator respectively.

3. Sudden load decrease in microgrid


The load in microgrid is decreased by cutting off two lines as shown in the figure below.

Figure 6.14. Sudden load decrease in microgrid.

27
The voltage, current, active power and reactive power simulation results at PCC are shown
below. Initially, the active power is being supplied from main grid to microgrid but after the
sudden decrease of load in microgrid, the flow of active power at PCC reverses. The reactive
power flow from main grid to microgrid drops down after the sudden load decrease in microgrid.
Also, the unbalance between the phases in the current at PCC increases after the load drop.

Figure 6.15. V, I, P and Q at PCC.


The voltage, current, active power and reactive power simulation results of battery are shown in
Figure 6.16. Battery is of comparatively lower power capacity and therefore, the output of
battery remains the same as before even after the load decrease.

Figure 6.16. V, I, P and Q of Battery.

28
The voltage, current, active power and reactive power simulation results of PV are shown below.
The PV active power steps up after the load cuts off. The reactive power supplied by PV first
increase for 0.2 seconds and then drops down.

Figure 6.17. V, I, P and Q of PV.

The voltage, current, active power and reactive power simulation results of Diesel Generator are
shown below. The reactive power supplied by diesel generator decreases after the load cuts off.
There is some variation in the output of the diesel generator due to the sudden load decreases but
it gradually become stable.

Figure 6.18. V, I, P and Q of Diesel Generator.

29
The frequencies at PCC, battery, PV and Diesel generator in the same order are shown below.
The frequencies become stable at 60 Hz after 0.05 seconds.

Figure 6.19. Frequencies at PCC, battery, PV and Diesel generator respectively.

Kundurs Two Area System

Kundurs Two Area System is then used to Kundurs two area system consists of two areas and
11 buses, connected by a weak tie between bus 7 and 9.The system operates at the fundamental
frequency of 60 Hz. Each area consists of two generators, each having a rating of 900 MVA and
20 kV. The load in Area 1 is 967MW +100MVAR -187MVAR with the shunt capacitance of
200MVAR. The load in Area 1 is 1767MW + 100MVAR - 187MVAR with the shunt
capacitance of 350MVAR. There is power flow of 413 MW from Area 1 to Area 2.

Figure 6.20. Tie-line diagram of Kundurs Two-Area System [21].

30
The Simulink model of this Kundurs Two-Area System is shown in Figure.6.20 (a),(b) and (c)
below. The microgrid is in Area 2.

Figure 6.21(a) Simulink Model of Kundurs Two-Area System

Figure6.21(b). Area 1 of Kundurs Two-Area System

Figure.6.21(2). Area 2 of Kundurs Two-Area System

31
The following contingency simulation was done to analyze the behavior of the system when the
microgrid is connected to the two-area system which emulates the behavior of generators in
large-scale real power grid system.

Three phase 10-cycles fault in grid-mode at Bus 1 followed by islanding:


The results obtained by simulating a three phase 10-cycles fault at Bus 1 in the two area system
shown in Figure 6.21(a) are given below.

In grid mode, active power at PCC is negative which means that grid is supplying power to
microgrid. The three-phase measurement used at PCC in this simulation is in reverse direction
compared to the one used in the previous simulations when an ideal voltage source was
connected as a grid. After disconnection from grid, active power of diesel generator increases
after islanding to provide for the load in microgrid as shown in figure 6.22.

Figure 6.22. Active Power Flow at PCC, Battery, PV and Diesel Generator.

The reactive power of diesel generator increases after islanding, reactive power of PC increases
for 0.5 seconds and then settles when reactive power of diesel generator starts providing the
required reactive power in microgrid as shown in Figure 6.23.

32
Figure 6.23. Reactive Power Flow at PCC, Battery, PV and Diesel Generator.
The frequencies of generators 1 and 2 closer to the fault location increases instantly after the
fault compared to the generators in Area 2which are farther away from the point of fault as
shown in figure below.

Figure 6.24. Frequencies of generators in Kundurs Two-Area System.


The frequencies of PCC, battery, PV and Diesel generator for this case of simulation are shown
in Figure 6.25.The frequencies of all 3 DGs dips to 58.3 Hz during fault but gets stable after
islanding.

33
Figure 6.25. Frequencies at PCC, battery, PV and Diesel generator.

The voltage, current, active power and reactive power simulation results of battery are shown
below. The reactive power output of the battery increases after islanding as shown. The voltage
drops and the current increase because there is no power flow from main grid to microgrid and
the distributed generators in microgrid provide for the load in microgrid.

Figure 6.26. V, I, P and Q of Battery

34
The voltage, current, active power and reactive power simulation results of PV are shown below.
The reactive power of PV increases after islanding to compensate the reactive load in the
microgrid.

Figure 6.27. V, I, P and Q of PV

The voltage, current, active power and reactive power simulation results of diesel generator are
shown below. The reactive power of diesel decreases for a short interval when reactive power of
PV is providing the reactive power to the microgrid.

Figure 6.28. V, I, P and Q of Diesel Generator in microgrid.

35
CHAPTER 7 : DECISION TREES

Overview and Applications

Data mining (also known as knowledge discovery) is the process of detecting underlying
correlations or patterns among different entities in large relational database. Data mining draws
ideas from machine learning, artificial intelligence, pattern recognition. Several data mining
techniques such as- Artificial Neural Networks, Decision Trees, Support Vector Machines and
Bayesian Networks have been used in the field of power system data analytics. Among these
data mining techniques, especially those with white box nature, such as artificial neural
network, support vector machine etc., decision trees has gained increasing interests because it not
only provides the insight information of data sets with low computational burden, but also
reveals the principles learnt by DTs for further interpretation. Apart from these features, decision
trees are better suited for managing the uncertainties associated with power systems through the
use of statistical methods to increase the reliability [17]. The following sections presents a brief
overview of decision tree based approaches in power system applications and provides an outline
for the decision tree based decision making approach adopted in this project.

Decision trees (DTs) belong to a class of data mining tools capable of extracting useful
information from large data set sand consequently provide assistance in decision-making
processes. It is a supervised learning system with the objectives of classification and prediction
similar to artificial neural networks and the group of other pattern recognition tools. In contrast
to many other techniques, the general DT methodology builds classifiers in a hierarchical form,
and selects automatically the most relevant features from among a (possibly large) list of
candidates. The resulting classifiers (the decision trees) have forms compatiblewith typical
human reasoning, which is easy to understand and interpret.

A DT-based method in general relies on three types of information: a learning set, composed of a
samples of solved cases (i.e. cases pre-classified with respect to a given criterion), a list of
critical attributes (i.e. of parameters likely to drive the phenomenaof concern, or in terms of
which it is desirable to characterizesecurity), a DT building method which will select out of this
listthe most relevant ones and formulate a security criterion in theform of a DT [18].

Decision trees have been previously used in electrical power system applications like- power
system security assessment, fault diagnosis, protection, forecasting and identification etc. Fig.7.1
presents a simple five node decision tree for power system security assessment purpose where
A and B are inputs to the decision tree, and the outcomes of the tree are classes SECURE
and INSECURE [19].

36
[A, B] A>K

A>K AK

B S SECURE

B S B

INSECURE SECURE
Figure 7.1. A simple 5 node decision tree [19]

A decision tree is made up of a set of nodes that uses supervised learning methods to achieve
classification or prediction for an objective variable. As shown in Fig.7.1, a DT can predict the
classification (e.g. SECURE or INSECURE) of an object. The object is represented by a
vector consisting of the values of a group of critical attributes (CAs, e.g. A and B in Fig.4.1). The
classification process consists of dropping the vector of CAs down the DT starting at the root
node until a terminal node is reached along a path, the class assigned to which is the
classification result. At each inner (nonterminal) node, a question (i.e., a splitting rule)
concerning a CA is asked to decide which child node the vector should drop into. For numerical
variable A, the question compares it with a threshold; for categorical variable B, the question
checks whether it belongs to a specified set.

Each of the elements (i.e. cases) in learning set and test set consists of a classification and a
vector comprisingof the values of a group of CA candidates (called predictors).The building
process initially grows a maximal tree by recursively splitting a set of learning cases(i.e., a parent
node) into two purer subsets (i.e., two new child nodes). To achieve each split, all possible
splitting rules relatedto predictors are scored by how well different classes of casesin the parent
node are separated. The splitting rule with the highest score isselected and called a critical
splitting rule (CSR). The othersplitting rules are called competitors. Some splitting rules that
can completely mimic the action of the CSR are called surrogates.Acompetitor with the same
improvement score as that ofthe CSR can generate an equivalently good split and a surrogatecan
generate exactly the same split as the CSR. The maximaltree is then pruned to generate a series
of smaller DTs. The testset is used to test their performance. A commonly used index isthe
misclassification cost, which is calculated using

(|) (7.1)

where, is the number of test cases, ( | ) is the cost of misclassifying a class case as a
class case and is the number of class cases whose predicted class is . The correctness
rate of classifying class cases is denoted by

( ) (7.2)

37
where, is the number of class test cases. For a sufficient number of test cases, a smaller
generally corresponds to a better decision tree. If statistical errors are also considered, the
standard error estimate for is denoted as and is calculated as following,


( ) (7.3)

Two decision trees whose cost difference is smaller than the standard error estimate of either one
have almost equal performance [19-20].

This project proposes a Decision Tree (DT) based approach for assisting microgrid central
controller in the decision making tasks under the following circumstances:

Islanding decision processes following a grid side contingency while the microgrid is
working in grid-connected mode.
Preventive and corrective control decision processes following contingencies within
microgrid while the microgrid system is working in islanding mode.

In this project, each of the above mentioned tasks are realized using the following steps:

Offline Contingency Database Preparation: A contingency database is prepared using


simulations studies performed on the microgrid model by recording all of the potential
critical attributes (CAs) and an offline classification. These databases are used to train
the decision trees.
Decision Tree Training: Decision trees are trained using the offline contingency
databases prepared in the previous step. Then the DTs are allowed to identify the critical
attributes (CAs) that characterize the decision criteria in the best possible way.
DT Testing and Performance Assessment: Test data sets for testing the trained DTs are
generated and performances of the decision trees are evaluated using the accuracy of the
classification.

38
Offline DT Building and Training
Generate a case database from contingency
simulations with classifications under the
security criteria specified.

Build a Decision Tree from the database

DT & Database

Periodic DT Update

Unpredicted OCs Exist?


No
Yes
Generate New Cases

Yes
DT Classifies New Cases Well?

No
Update the database and rebuild DT

Final DT

Online Assessment
Compare the test data and
measurements with the thresholds
stored in the DT

Assessment Results

Fig.7.2. DT based scheme for microgrid islanding and controls

In the above mentioned procedure, each DT is built using J48 decision tree algorithm which is
the Java version of the popular C4.5 decision tree algorithm. Fig.7.2 presents the conceptual
flowchart for the decision tree building approach used in this project from database generation,
training and testing phases.

39
Decision Tree for Microgrid Islanding Assistance

Security Criteria, Implementation Methodology, Test Results


The ever increasing data availability in modern power systems prompts the application of data
mining tools for power system data analysis. Prominent data mining techniques like decision tree
have been utilized for power system security assessment, preventive and corrective controls,
fault diagnosis, protection, forecasting and identification. This project intends to utilize decision
trees in assisting microgrid central controller in islanding decision making while working in the
grid connected mode as well as in providing assistance in preventive and corrective control
decisions while working in the islanding mode.

Microgrid central controller is required to island the system from main grid following system
contingencies that violates the standard interconnection requirements for distributed energy
resources set by IEEE Standard 1547. According to IEEE 1547, the islanding decisions while in
grid connected mode are predominantly based on system voltage and frequency measurements at
point of common coupling. Decision trees can be trained and implemented to detect underlying
patterns and relationships among various system variables that can be used for predicting system
voltages and frequency following contingencies for assisting the islanding decision process.

The following subsections provide a brief overview of the islanding decision criteria based on
system voltage and frequency requirements from IEEE 1547 and summarizes the implementation
of decision trees for assistance in islanding decision including the event database creation, data
preprocessing, tree structures and preliminary results.

Review of IEEE 1547 Requirements on Voltage


The protection functions of the interconnection system shall detect the effective (RMS) or
fundamental frequency value of each phase-to-phase voltage, except where the transformer
connecting the DR to the Area EPS is a grounded wye-wye configuration, or single-phase
installation, the phase-to-neutral voltage shall be detected. When any voltage is in a range given
in Table.4.2, the DR shall cease to energize the Area EPS within the clearing time as indicated.
Clearing time is the time between the start of the abnormal condition and the DR ceasing to
energize the Area EPS. For DR less than or equal to 30 kW in peak capacity, the voltage set
points and clearing times shall be either fixed or field adjustable. For DR greater than 30 kW, the
voltage set points shall be field adjustable [15].
The voltages shall be detected at either the PCC or the point of DR connection when any of the
following conditions exist:

1. The aggregate capacity of DR systems connected to a single PCC is less than or equal
to 30 kW.
2. The interconnection equipment is certified to pass a non-islanding test for the system
to which it is to be connected.
3. The aggregate DR capacity is less than 50% of the total Local EPS minimum annual
integrated electrical demand for a 15-minute time period, and export of real or reactive power by
the DR to the Area EPS is not permitted [15].

40
Table 7.1. Interconnection System Response to Abnormal Voltages [15]

Voltage Range Clearing time


[% of base voltage] [sec]
V < 50 0.16
50 V < 88 2.00
110 V < 120 1.00
V 120 0.16

Review of IEEE 1547 Requirements on Frequency:


When the system frequency is in a range given in Table 4.3, the DR shall cease to energize the
Area EPS within the clearing time as indicated. Clearing time is the time between the start of the
abnormal condition and the DR ceasing to energize the Area EPS. For DR less than or equal to
30 kW in peak capacity, the frequency set points and clearing times shall be either fixed or field
adjustable. For DR greater than 30 kW the frequency set points shall be field adjustable.
Adjustable under frequency trip settings shall be coordinated with Area EPS operations [15].

Table 7.2. Interconnection System Response to Abnormal Frequencies [15]


Frequency Range Clearing Time
DR Capacity
[Hz] [sec]
>60.5 0.16
30 KW
<59.3 0.16
>60.5 0.16
>57 and <59.8
>30KW Adjustable 0.16 to 300
Adjustable set point
<57.0 0.16

Data Preprocessing and Decision Tree Training:


Decision trees are trained using contingency databases created from fault simulations on the grid
side in the simulation model. The simulation model consists of a two area electrical power
system with the microgrid connected in bus B2 in area-2 of the system. Two contingency
databases have been created with identical grid side fault simulation scenarios for classification
of voltage and frequency abnormalities. Each of these databases contain 48 cases generated from
6 buses from Area-1 and Area-2 under 4 different types of faults and two different system
loading scenarios.

The training data set for decision trees have been preprocessed to include the potential critical
attributes that can characterize and predict the events related to voltage and frequency
abnormalities as well as sample voltage and frequency classifications have been generated for
the supervised training of the decision trees.
The events associated with voltage and frequency abnormalities mentioned in IEEE Standard
1547 are classified into five subcategories for voltage related events and four subcategories for

41
frequency related events. Based on these classification results, each of the fault simulation cases
have been labeled with frequency and voltage classes which have been used to train the decision
tree. Table 7.3 and Table 7.4 present the voltage and frequency classes along with the IEEE 1547
standards respectively.

Table 7.3. Interconnection System Response to Abnormal Voltages with Class Identifier

Voltage Range Clearing time Class Identifier


[% of base voltage] [sec]
V < 50 0.16 VU2
50 V < 88 2.00 VU1
88 V < 110 - VN
110 V < 120 1.00 VO1
V 120 0.16 VO2

Table 7.4. Interconnection System Response to Abnormal Frequencies with Class Identifier
Frequency Range Clearing Time Class Identifier
DR Capacity
[Hz] [sec]
>60.5 0.16 FO1
30 KW 59.8< f < 60.5 - FN
<59.3 0.16 FU1
>60.5 0.16 FO1
59.8< f < 60.5 - FN
>30KW >57 and <59.8
Adjustable 0.16 to 300 FU1
Adjustable set point
<57.0 0.16 FU2

Table 7.5 lists the critical attributes used in developing the decision trees for voltage and
frequency classification purpose in order to assist in the islanding decision process. The list
includes voltage, current, frequency, real and reactive power values. In addition to those system
parameters, a series of abnormal voltage and frequency event identifiers have also been included
in the list. The elements in the critical attribute list have been chosen keeping in mind the
possibility of losing a certain type of measurements (i.e. voltage measurements) and yet being
able to predict and classify the incoming voltage and frequency abnormalities. The voltage and
frequency events identifier included in the list can be utilized not only to detect the abnormal
events but also they can be utilized in detecting the sequence of the different classes of voltage
and frequency abnormalities resulting from the same grid side contingency.

Table 7.6 and Table 7.7represents the preprocessed training data for decision tree training for
frequency and voltage classification problems respectively. Table 7.8 and 7.9 presents the
ranking of the predictors for the classification tasks. The ranking is obtained by removing the
best ranked predictor from the training dataset in each of the previous iterations during the
training process and allowing the decision tree algorithm to select the most suitable and relevant
predictor in the current iteration. The corresponding accuracies are obtained under four different

42
test scenarios. The first sets of accuracies are obtained using the training data as the testing
dataset. The second set of accuracies are obtained using the 10 folds Cross-Validation
approach which basically divides the entire dataset into 10 individual pieces, and uses one of the
pieces for testing the decision tree and the remaining nine pieces are used for training the
decision tree. In this way the algorithm performs 11 iterations with each of the pieces in the
dataset is used at least once as the testing data in the first 10 iterations and in the final iteration
the entire dataset is treated as the testing data. From each of the iterations, accuracies are
calculated and their average is computed which is the final accuracy result. This approach
provides somewhat more realistic accuracy results compared to using the training dataset for
testing the decision trees. The third and fourth set of accuracies are computed using 66% of the
training dataset to train the tree and the remaining 34% to test the decision tree, and with 85% of
the training dataset to train the tree and the remaining 15% of the dataset to test the decision tree
respectively.

43
Table 7.5. List of critical attributes selected for building the decision trees
Critical Attribute Category Unit Description
Fault Type(3p/1p) Nominal - Type of the fault on the grid side
Location Nominal - Location of the fault (Bus Number)
Area Nominal - Area of the Faulted Bus
Minimum phase-neutral voltage in
Vamin Numeric kV (rms)
phase-A
Maximum phase-neutral voltage in
Vamax Numeric kV (rms)
phase-A
Minimum phase-neutral voltage in
Vbmin Numeric kV (rms)
phase-B
Maximum phase-neutral voltage in
Vbmax Numeric kV (rms)
phase-B
Minimum phase-neutral voltage in
Vcmin Numeric kV (rms)
phase-C
Maximum phase-neutral voltage in
Vcmax Numeric kV (rms)
phase-C
Iamin Numeric A (rms) Minimum phase current in phase-A
Iamax Numeric A (rms) Maximum phase current in phase-A
Ibmin Numeric A (rms) Minimum phase current in phase-B
Ibmax Numeric A (rms) Maximum phase current in phase-B
Icmin Numeric A (rms) Minimum phase current in phase-C
Icmax Numeric A (rms) Maximum phase current in phase-C
Pmin Numeric MW Minimum real power
Pmax Numeric MW Maximum real power
Qmin Numeric MVAR Minimum real power
Qmax Numeric MVAR Maximum real power
Fmin Numeric Hz Minimum frequency
Fmax Numeric Hz Maximum frequency
VU2a Numeric - Class identifier
VU1a Numeric - Class identifier
VO1a Numeric - Class identifier
VO2a Numeric - Class identifier
VU2b Numeric - Class identifier
VU1b Numeric - Class identifier
VO1b Numeric - Class identifier
VO2b Numeric - Class identifier
VU2c Numeric - Class identifier
VU1c Numeric - Class identifier
VO1c Numeric - Class identifier
VO2c Numeric - Class identifier
FO1 Numeric - Class identifier
FU1 Numeric - Class identifier
FU2 Numeric - Class identifier

44
Table 7.6. Decision tree Training Data for Frequency Classification Problem

45
Table 7.7. Decision tree Training Data for Voltage Classification Problem

46
Table 7.8. Ranking of Predictor Variables and Corresponding Accuracy in Frequency Classification Problem
Rank Predictor Accuracy [Training Accuracy [Cross Accuracy [Training Accuracy [Training 85%,
Data used for Testing] Validation, 10 Folds] 66%, Testing 34% ] Testing 15% ]
1 Fmin 100 91.9167 100 100
2 Fmax, Ibmax, Area 97.9167 85.4167 75 85.7143
3 Icmax, Vamin, Fault 97.9167 87.5 75 100
Type
4 Qmin, Area, Ibmax 97.9167 85.4167 75 85.7143
5 Vamax, Pmin, Pmax 97.9167 81.25 75 85.7143
6 Vbmax, Pmin, Pmax 97.9167 81.25 75 100
7 Vcmax, Pmin, Pmax 97.9167 83.33 75 100
8 Pmin, Fault Type, Vamin 100 85.4167 75 57.1429
9 Icmin, VO2b, Qmax, 100 83.333 75 57.1429
Fault Type
10 Qmax, Area, Iamax 95.833 89.5833 75 57.1429
11 Location, VO2c, Fault 95.833 89.5833 75 57.1429
Type
12 Area, Vamin, Iamax 95.833 85.4167 75 57.1429
13 Fault Type 87.5 81.25 75 57.1429
14 VO2b, Vamin, Pmax, 100 70.833 62.5 71.4286
VU2a, Vbmin, Iamin
15 Vamin, Pmax, VU2a, 100 79.1667 56.25 85.7143
Vbmin, Iamin
16 Vbmin, Pmax, VU2a, 100 77.0833 56.25 85.7143
Vcmin, Vbmin, Iamin
17 Vcmin, Pmax 91.667 77.0833 62.5 85.7143
18 Ibmax, Iamin, VU2a 95.833 85.4167 56.25 100
19 Iamax, VU2a, Iamin 97.9167 83.333 56.25 100
20 Iamin, VO2c 89.5833 64.5834 56.25 71.4286
21 Ibmin, VU2c, Pmax, 95.8333 70.8333 56.25 71.4286
VU2a
22 Pmax, VO2a 79.1667 56.25 56.25 71.4286

47
Table 7.9. Ranking of Predictor Variables and Corresponding Accuracy in Voltage Classification Problem
Rank Predictor Accuracy [Training Accuracy [Cross Accuracy [Training Accuracy [Training
Data used for Testing] Validation, 10 Folds] 66%, Testing 34% ] 85%, Testing 15% ]
1 VU2a 100 100 100 100
2 Vamax 100 91.6667 87.5 85.7143
4 Vcmax 100 91.6667 87.5 85.7143
5 VU2b, VO2b 100 91.6667 87.5 85.7143
6 VO2c, VO2b 100 95.833 100 100
7 Vbmin, Vbmax 100 95.833 100 100
8 VO2a, VU2c 100 97.9167 100 100
9 VU2c, Icmax, Area 100 87.5 75 100
10 VO2b, Icmax, Area 100 93.75 93.75 100
11 Vbmax, Icmin 100 93.75 93.75 100
12 Icmin, Vamin 100 95.833 87.5 100
13 Vamin, Area 100 93.75 87.5 100
14 Area, Vcmin 100 89.5833 93.75 100
15 Icmax, Ibmin, Location, 100 81.25 93.75 85.7143
Iamin
16 Vcmin, Location, Iamin 100 89.5833 87.5 85.7143
17 Iamin, Pmin, Pmax 100 93.75 87.5 100
18 Ibmin, Pmin, Pmax 100 91.667 87.5 100
19 Location, Ibmax, Iamax, 93.75 77.0833 75 85.7143
Qmax
20 Pmax, Pmin, Fault Type 100 66.6667 75 71.4286
21 Fault type, Iamax, Pmin, 100 66.6667 75 71.4286
Ibmax
22 Pmin, Qmin, Ibmax 100 85.4167 75 85.7143
23 Qmin, Ibmax, Iamax 100 72.9167 68.75 85.7143
24 Qmax, Iamax 97.9167 79.1667 68.75 85.7143

48
Fig.7.3-Fig.7.10 presents some of the sample decision tree structures obtained from the analysis
so far along with their associated predictor variables and the corresponding accuracies from the
four methods mentioned above.

Fig.7.3 represents the application of current measurements in detecting the violations in voltage
standards mentioned in IEEE 1547. Fig.7.4 represents a combination of real and reactive power
measurements along with current measurements to predict the voltage standard violations. The
figures also include the accuracy values from the four different methods used to test the
developed decision trees along with the predictors used. The maximum accuracy among the
different testing methods (cross validation, percent split- 66, 34 and percent split-85, 15) is 100%
and the minimum is 70.83%. The accuracy values indicates that these combinations can be
utilized to predict the voltage violation events even though the voltage measurements are not
available for the classification and prediction purpose.Fig.7.5 and Fig.7.6 presents the utilization
of combinations of voltage and current measurements, and voltage and power measurements in
detecting the abnormal frequency events. The maximum accuracy among the different testing
methods (cross validation, percent split- 66, 34 and percent split-85, 15) is 83.33% and the
minimum is 67.55%. The accuracy values presents the possibilities of utilizing the voltage and
power measurements in detecting frequency abnormalities with reasonable accuracies.Fig.7.7
and Fig.7.8 presents the use of combination of voltages, currents and power measurements along
with fault types to predict the frequency abnormalities for microgrid islanding purpose. These
cases include training data consisting of 24 different contingencies under two different loading
conditions in the microgrid. The maximum and minimum accuracies are 100% and 75%,
respectively from the three different testing approaches employed.
Fig.7.9 and Fig.7.10 shows the use of current and power measurements in classification of
voltage abnormalities. These cases also include training data consisting of 24 different
contingencies under two different loading conditions in the microgrid. The maximum and
minimum accuracies are 100% and 66.67%, respectively from the three different testing
approaches employed.

These experiments indicates the abilities of decision trees in classification and prediction of
relevant events even when the most relevant or important measurements are not available. There
might be a compromise in the classification and prediction accuracies, but the accuracies in the
experiments indicates that the decision trees still achieve reasonable accuracies and these values
can be further improved with the enhancements (i.e. including more contingency cases) of the
training dataset.

49
Predictors Ibmin, Icmin, Ibmax Predictors Ibmax, Icmin
Accuracy(training data) 100 % Accuracy(training data) 100 %
Accuracy(Cross Validation) 70.833% Accuracy(Cross Validation) 87.5%
Accuracy (% Split-66,34) 100% Accuracy (% Split-66,34) 100%
Accuracy (% Split-85,15) 100% Accuracy (% Split-85,15) 100%

(a) (b)
Figure 7.3. Decision trees for voltage classification: (a) with predictors Ibmin, Icmin, Ibmax; (b) with predictors
Ibmax and Icmin

Predictors Icmin, Pmax, Qmax Predictors Qmax, Iamax


Accuracy(training data) 95.83 % Accuracy(training data) 95.83%
Accuracy(Cross Validation) 79.17% Accuracy(Cross Validation) 83.33%
Accuracy (% Split-66,34) 100% Accuracy (% Split-66,34) 100%
Accuracy (% Split-85,15) 100% Accuracy (% Split-85,15) 100%
(a) (b)
Figure 7.4. Decision trees for voltage classification: (a) with predictors Icmin, Pmax and Qmax; (b) with predictors
Qmax and Iamax

50
Predictors Icmin, Vamin Predictors Iamax, Icmin
Accuracy(training data) 100 % Accuracy(training data) 95.83 %
Accuracy(Cross Validation) 83.33% Accuracy(Cross Validation) 79.17%
Accuracy (% Split-66,34) 75% Accuracy (% Split-66,34) 75%
Accuracy (% Split-85,15) 75% Accuracy (% Split-85,15) 72%
(a) (b)
Figure 7.5. Decision trees for frequency classification: (a) with predictors Icmin and Vamin; (b) with predictors
Iamax and Icmin

Predictors Qmax, Vamin Predictors Vcmax, Vbmin


Accuracy(training data) 95.83 % Accuracy(training data) 95.83 %
Accuracy(Cross Validation) 79.17% Accuracy(Cross Validation) 70.833%
Accuracy (% Split-66,34) 75% Accuracy (% Split-66,34) 75%
Accuracy (% Split-85,15) 67.55% Accuracy (% Split-85,15) 67.55%
(a) (b)
Figure 7.6. Decision trees for frequency classification: (a) with predictors Qmax and Vamin; (b) with predictors
Vcmax and Vbmin

51
Predictors Icmax, Vamin, Fault Type
Accuracy(training data) 97.92 %
Accuracy(Cross Validation) 87.50%
Accuracy (% Split-66,34) 75%
Accuracy (% Split-85,15) 100%

Figure 7.7. Decision trees for frequency classification with predictors Icmax, Vamin and Fault types

Predictors Vamax, Pmin, Pmax


Accuracy(training data) 97.92 %
Accuracy(Cross Validation) 81.25%
Accuracy (% Split-66,34) 75%
Accuracy (% Split-85,15) 85.72%

Figure 7.8. Decision trees for frequency classification with predictors Vamax, Pmin and Pmax

52
Predictors Iamin, Pmin, Pmax
Accuracy(training data) 100 %
Accuracy(Cross Validation) 93.75%
Accuracy (% Split-66,34) 87.5%
Accuracy (% Split-85,15) 100%

Figure 7.9. Decision trees for voltage classification with predictors Iamin, Pmin and Pmax

Predictors Fault Type, Iamax, Pmin, Ibmax


Accuracy(training data) 100 %
Accuracy(Cross Validation) 66.67%
Accuracy (% Split-66,34) 75%
Accuracy (% Split-85,15) 71.43%

Figure 7.10. Decision trees for voltage classification with predictors Fault types, Iamax, Pmin and Ibmax

53
CHAPTER 8
CONCLUSION AND RECOMMENDATIONS

The microgrid will play a very vital role in the future grid-system and the penetration of
distributed generators will increase more than ever. Therefore, it is very important to study the
impact of this distributed generation penetration for ensuring a more reliable and efficient power
grid system. In this thesis work, three distributed generators were connected to form a microgrid
capable of stable operation in both grid-connected and islanding mode. All three DGs model
developed in Simulink MATLAB satisfy the current harmonics and DC injection standards set
by IEEE 1547 Standard for interconnection with the main grid system. Battery control was
designed for load shaving during peak hours of load when the load increases from the planned
load. The control strategies in these DGs is developed in such a way that the microgrid is capable
of functioning properly in grid-mode and also, capable of independently working in islanding-
mode. In grid-connected mode, all DGs follow the frequency of the main grid whereas in
islanding-mode, the diesel generator provides the reference frequency which the other DGs
follow. Various scenarios of possible contingencies in this microgrid system in both grid-
connected and islanding mode were studied. The microgrid was connected to Kundurs two area
system to better emulate the behavior of the main power system. The results were analyzed to
better understand the behavior of DGs in the microgrid as well as the generators in the Kundurs
two-area system. Four cases of such contingency simulations are discussed in this thesis work.
The study of these contingencies gives a better understanding of the behavior of distributed
generators.

To further this study, another distributed generator e.g. wind turbine can be added to this system
to further study the behavior of microgrid and main grid under different scenarios. Also, since
the load in IEEE 13 Node Test Feeder System is unbalanced therefore negative sequence
compensators can be implemented in the power electronics converter of any DG to improve the
voltage and current waveforms in the microgrid.

For the purpose of developing a decision trees based protection and control framework or fault-
detection algorithm for microgrids against disturbances, various cases of contingencies in
external grid were simulated with the microgrid in grid-connected mode. The data was collect at
PCC and from this data, a list of critical attributes was developed which includes the parameter
likely to be of concern in terms of security etc. Weka software is a collection of machine learning
algorithms for data mining tasks. Weka contains tools for data pre-processing, classification,
regression, clustering, association rules, and visualization. It is also well-suited for developing
new machine learning schemes. Decision trees are trained to detect underlying patterns and
relationships among various system parameter which can be used for predicting system voltages
and frequency following contingencies for assisting the islanding decision process. The decision
trees using this software point out to the parameters which are likely to predict the violation of
any security criteria for islanding. Fault information and voltage/frequency measurements at the
PCC bus have the highest rankings. If fault information or voltage/frequency measurements are
not available, current, real power or reactive power measurements can be alternatives to help
make the decision.
54
The 48 cases simulations and decision tree development done in this thesis work show the
potential of decision trees in classification and prediction of concerned events even when the
most relevant or important measurements are not available. There may be some compromises in
the classification and prediction accuracies, but the results show that they are still reasonable and
can be made more reliable by adding more possible cases. A bigger database can thus further be
developed to build, refine and improve the decision trees. This larger database can be used to
include all the possible contingencies and scenarios to make these decision trees more and more
reliable and practical.

55
LIST OF REFERENCES

56
[1] The World Alliance for Decentralized Energy (WADE), World Survey on De-
centralized Energy 2006, 46 pages, May 2006 (source: http://www.localpower.org )
[2] R. H. Lasseter, et al., "White paper on integration of distributed energy resources, The
CERTS microgrid concept. Consortium for Electric Reliability Technology Solutions,
pp. 1-27, 2002.
[3] L. Blackburn, T. J. Domin, Protective relaying Principles and Applications, Third
Edition, CRC Press, 2006.
[4] M. R. Islam, H. A. Gabbar, Analysis of Microgrid Protection Strategies, IEEE
International Conference on Smart Grid Engineering, August, 2012.
[5] C. Marnay, F. Rubio, and A. Siddiqui, "Shape of the microgrid", IEEE Power
Engineering Society Winter Meeting, Vol. 1, pp.150 - 153, 2001.
[6] IEEE 13 Node Test Feeder System,(source:
http://ewh.ieee.org/soc/pes/dsacom/testfeeders/ )
[7] "Commercial Opportunities for Back-Up Generation and Load Reduction via National Grid,
the National Electricity Transmission System Operator (NETSO) for England, Scotland,
Wales and Offshore," 2009-12-22. Retrieved 2012-07-13(source: www.claverton-
energy.com).
[8] U. Gangopadhyay, S. Jana, and S. Das, State of Art of Photovoltaic Technology,
Conference Papers in Energy, Volume 2013 (2013), Article ID 764132, 9 pages, (source:
http://dx.doi.org/10.1155/2013/764132 ).
[9] R.N.Tripathi, A. Singh, M.Badoni, A MATLAB Simulink based Solar Photovoltaic
Array SPVA module with MPPT,Emerging Trends in Communication, Control, Signal
Processing & Computing Applications (C2SPCA), 2013 International Conference onOct.
2013.
[10] O. Tremblay, L. Dessaint,Experimental Validation of a Battery Dynamic Model for EV
Applications, EVS24 Stavanger, Norway, May 13 - 16, 2009
[11] Y.Iijima, Y. Sakanaka, N. Kawakami, M. Fukhuhara, Development and field
experiences of NAS battery inverter for power stabilization of a 51 MW wind farm,
Power Electronics Conference (IPEC), 2010 International, Page(s): 1837-1841.
[12] Y.Liu, Improvement of Available Battery Capacity,Journal of Power
Electronics, Vol. 13, No. 3, May 2013.
[13] H. Bai, C. Mai, The impact of bidirectional DC-DC converter on the inverter
operation and battery current in hybrid electric vehicles,Power Electronics and ECCE
Asia (ICPE & ECCE), 2011 IEEE 8th International Conference onMay 2011-June2011,
Page(s): 1013-1015.
[14] M.A Sofla, L. Wang,Control of DC-DC Bidirectional Converters for Interfacing
Batteries in Microgrid,Power Systems Conference and Exposition (PSCE), 2011
IEEE/PES on March 2011, Page(s): 1-6.
[15] IEEE Application Guide for IEEE Std. 1547- IEEE Standard for Interconnecting
Distributed Resources with Electric Power Systems, IEEE Std. 1547-2003, Jun. 2003.
[16] Y.Wang, Z. Lu, Y. Min, "Analysis and comparison on the control strategies of multiple
voltage source converters in autonomous microgrid", Developments in Power System
Protection (DPSP 2010). Managing the Change, 10th IET International Conference, 2010
, Page(s): 1- 5.
57
[17] C. Liu, Z.H. Rather, Z. Chen, C. L. Bak, An Overview of Decision Tree Applied to
Power System, Intl. Journal of Smart Grid and Clean Energy, vol.2, no.3, pp.414-419,
Oct. 2013.
[18] L. Wenhenkel, M. Pavella, Advances in Decision Trees Applied to Power System
Security Assessment, in Proc. IEE Advances in Power System Control, Operation and
Management, Dec. 1993, pp.47-53
[19] K. Sun, S. Likhate, V. Vittal, V.S. Kolluri, S. Mandal, An Online Dynamic Security
Assessment Scheme Using Phasor Measurements and Decision Trees, IEEE Trans. on
Power Systems, vol.22, no.4, pp.1935-1943, Nov. 2007.
[20] K. E. Arroudi, G. Joos, Data Mining Approach to Threshold Setting of Islanding Relays
in Distributed Generation, IEEE Trans. on Power Systems, vol.22, no.3, pp.1112-1119,
Aug. 2007.
[21] P.Kundur Author, Power System Stability and Control, 1994.

[22] C. Tammineedi, Modeling battery-ultracapacitor hybrid system for solar and wind
application, 2011 Thesis , Available: https://etda.libraries.psu.edu/paper/11561/6886

58
VITA
Hira Amna Saleem received her B.Sc. degree in Electronics engineering from Ghulam Ishaq
Khan Institute of Science and Technology, Pakistan, in 2010. This thesis is the work done for the
partial fulfillment of her M.Sc degree in Electrical Engineering at the University of Tennessee,
Knoxville.

59